I was recently at an acrobatics festival as the event photographer, and over the weekend I took almost 7000 photos. Most of these are "bursts" of 3-10 photos as I would just press-hold the shutter. Normally I do my photo review in darktable, but it takes a second or so to load a photo. This is very frustrating when trying to pick the best photo of a long burst. So what was I to do? Well, I spent a few hours with an AI and built myself a photo review app.
AI is an amazing tool. In the past if I had a project idea I'd have to prioritize it against other ideas, and then maybe spend a few weeks building it. These days if the idea is in certain domains, I can just ask the AI to do it. "Make me a photo review app" isn't likely going to build what I want. AI isn't magic, it's not going to know what I want. Instead I have to describe it: group photos by bursts, rating between 1-5, store ratings in xmp files, etc. Giving it one super mega prompt isn't the way either. Step it through - set up an application, load photos etc.
And then, with probably around an hour or two of active time, and only a few lines of actually writing code an application is made: it's not perfect, I wouldn't ship it to customers, but it gets the job done - and that's the point. In line with this post I think that as the cost of developing software drops, the amount of "I'll just ask the computer to do this" will increase, to the point where it is mostly seamless.
So it's not that one day AI will mantain all legacy software, it's that one day there'll be no point in having some super huge application. It won't be that vibe coding takes over. You just ask the computer to do it, and it'll make it happen. We'll call it Tuesday.